Completely Derandomized Self-Adaptation in Evolution Strategies (CMA-ES)

Adapted the full covariance matrix of a multivariate normal mutation distribution, achieving invariance to order-preserving objective transformations and rotational coordinate shifts.

Serves as qualified theoretical and systems foundation for task-black-box-optimization, task-hyperparameter-tuning.

Assumptions

• Markovian state dynamics and stationary reward functions hold in target evaluation environments

Limitations

• Sample efficiency, exploration stability, and real-world sim-to-real transfer gaps require specialized tuning